1. Field of the Invention
The present invention relates to switches, and particularly to a switch circuit and method of switching radio frequency (RF) signals within an integrated circuit. In one embodiment, the switch circuit comprises CMOS devices implemented on a silicon-on-insulator (SOI) substrate, for use in RF applications such as wireless communications, satellites, and cable television.
2. Description of Related Art
As is well known, radio frequency (RF) switches are important building blocks in many wireless communication systems. RF switches are found in many different communications devices such as cellular telephones, wireless pagers, wireless infrastructure equipment, satellite communications equipment, and cable television equipment. As is well known, the performance of RF switches is controlled by three primary operating performance parameters: insertion loss, switch isolation, and the xe2x80x9c1 dB compression point.xe2x80x9d These three performance parameters are tightly coupled, and any one parameter can be emphasized in the design of RF switch components at the expense of others. A fourth performance parameter that is occasionally considered in the design of RF switches is commonly referred to as the switching time or switching speed (defined as the time required to turn one side of a switch on and turn the other side off). Other characteristics that are important in RF switch design include ease and degree (or level) of integration of the RF switch, complexity, yield, return loss and cost of manufacture.
These RF switch performance parameters can be more readily described with reference to a prior art RF switch design shown in the simplified circuit schematics of FIGS. 1a-1c. FIG. 1a shows a simplified circuit diagram of a prior art single pole, single throw (SPST) RF switch 10. The prior art SPST switch 10 includes a switching transistor M15 and a shunting transistor M27. Referring now to FIG. 1a, depending upon the state of the control voltages of the two MOSFET transistors M15 and M27 (i.e., depending upon the DC bias applied to the gate inputs of the MOSFET switching and shunting transistors, M1 and M2, respectively), RF signals are either routed from an RF input node 1 to, an RF output node 3, or shunted to ground through the shunting transistor M27. Actual values of the DC bias voltages depend upon the polarity and thresholds of the MOSFET transistors M15 and M27. Resistor R09, in series with the RF source signal, isolates the bias from the source signal and is essential for optimal switch performance. FIG. 1b shows the xe2x80x9conxe2x80x9d state of the RF switch 10 of FIG. 1a (i.e., FIG. 1b shows the equivalent small-signal values of the transistors M1 and M2 when the RF switch 10 is xe2x80x9conxe2x80x9d, with switching transistor M15 on, and shunting transistor M27 off). FIG. 1c shows the xe2x80x9coffxe2x80x9d state of the switch 10 of FIG. 1a (i.e., FIG. 1c shows the equivalent small-signal values of the transistors M1 and M2 when the RF switch 10 is xe2x80x9coffxe2x80x9d, with switching transistor M15 off, and shunting transistor M27 on).
As shown in FIG. 1b, when the RF switch 10 is on, the switching transistor M15 is primarily resistive while the shunting transistor M27 is primarily capacitive. The xe2x80x9cinsertion lossxe2x80x9d of the RF switch 10 is determined from the difference between the maximum available power at the input node 1 and the power that is delivered to a load 11 at the output node 3. At low frequencies, any power lost is due to the finite on resistance xe2x80x9crxe2x80x9d 13 of the switching transistor M15 when the switch 10 is on (see FIG. 1b). The on resistance r 13 (FIG. 1b) typically is much less than the source resistor R09. The insertion loss, xe2x80x9cILxe2x80x9d, can therefore be characterized in accordance with Equation 1 shown below:
IL is approximately equal to: 10r/R0 ln(10)=0.087 r (in dB).xe2x80x83xe2x80x83Equation 1
Thus, at low frequencies, a 3-xcexa9 value for r results in approximately 0.25 dB insertion loss. Because insertion loss depends greatly upon the on resistances of the RF switch transmitters, lowering the transistor on resistances and reducing the parasitic substrate resistances can achieve improvements in insertion loss.
In general, the input-to-output isolation (or more simply, the switch isolation) of an RF switch is determined by measuring the amount of power that xe2x80x9cbleedsxe2x80x9d from the input port into the output port when the transistor connecting the two ports is off. The isolation characteristic measures how well the RF switch turns off (i.e., how well the switch blocks the input signal from the output). More specifically, and referring now to the xe2x80x9coffxe2x80x9d state of the RF switch 10 of FIG. 1c, the switching transistor M15 off state acts to block the input 1 from the output 3. The shunting transistor M27 also serves to increase the input-to-output isolation of the switch 10.
When turned off (i.e., when the RF switch 10 and the switching transistor M15 are turned off), M15 is primarily capacitive with xe2x80x9cfeedthroughxe2x80x9d (i.e., passing of the RF input signal from the input node 1 to the output node 3) of the input signal determined by the series/parallel values of the capacitors CGD off 15 (Gate-to-Drain Capacitance when the switching transistor M1 is turned off), CGS off 17 (Gate-to-Source Capacitance when the switching transistor M1 is turned off), and CDS119 (Drain-to-Source capacitance when the transistor M1 is turned off). Feedthrough of the input signal is undesirable and is directly related to the input-to-output isolation of the RF switch 10. The shunting transistor M27 is used to reduce the magnitude of the feedthrough and thereby increase the isolation characteristic of the RF switch.
The shunting transistor M27 of FIG. 1c is turned on when the switching transistor M15 is turned off. In this condition, the shunting transistor M27 acts primarily as a resistor having a value of r. By design, the value of r is much less than the characteristic impedance of the RF source. Consequently, r greatly reduces the voltage at the input of the switching transistor M15. When the value of r is much less than the source resistance R09 and the feedthrough capacitive resistance of the shunting transistor M27, isolation is easily calculated. Switch isolation for the off state of the RF switch 10 is determined as the difference between the maximum available power at the input to the power at the output.
In addition to RF switch insertion loss and isolation, another important RF switch performance characteristic is the ability to handle large input power when the switch is turned on to ensure that insertion loss is not a function of power at a fixed frequency. Many applications require that the switch does not distort power transmitted through a xe2x80x9cswitched-onxe2x80x9d switch. For example, if two closely spaced tones are concurrently passed through an RF switch, nonlinearities in the switch can produce inter-modulation (IM) and can thereby create a false tone in adjacent channels. If these adjacent channels are reserved, for instance, for information signals, power in these false tones must be maintained as small as possible. The switch compression, or xe2x80x9c1 dB compression pointxe2x80x9d (xe2x80x9cP1 dBxe2x80x9d), is indicative of the switch""s ability to handle power. The P1 dB is defined as the input power at which the insertion loss has increased by 1 dB from its low-power value. Or stated in another way, the 1 dB compression point is a measure of the amount of power that can be input to the RF switch at the input port before the output power deviates from a linear relationship with the input power by 1 dB.
Switch compression occurs in one of two ways. To understand how switch compressing occurs, operation of the MOSFET transistors shown in the RF switch 10 of FIGS. 1a-1c are described. As is well known in the transistor design arts, MOSFETs require a gate-to-source bias that exceeds a threshold voltage, Vt, to turn on. Similarly, the gate-to-source bias must be less than Vt for the switch to be off. Vt is positive for xe2x80x9ctype-Nxe2x80x9d MOSFETs and negative for xe2x80x9ctype-Pxe2x80x9d MOSFETs. Type-N MOSFETs were chosen for the RF switch 10 of FIGS. 1a-1c. The source of a type-N MOSFET is the node with the lowest potential.
Referring again to FIG. 1c, if a transient voltage on the shunting transistor M27 results in turning on the shunting transistor M27 during part of an input signal cycle, input power will be routed to ground and lost to the output. This loss of power increases for increased input power (i.e., input signals of increased power), and thereby causes a first type of compression. The 1 dB compression point in the RF switch 10 is determined by the signal swing on the input at which point the turned-off shunting transistor M27 is unable to remain off. Eventually, a negative swing of the input falls below the potential of the M2 gate, as well as below ground (thus becoming the source). When this difference becomes equal to Vt, the transistor M27 begins to turn on and compression begins. This first type of compression is caused by the phenomenon of the turning on of a normally off gate in the shunt leg of the RF switch. Once the shunting transistor M27 turns on, power at the output node 3 no longer follows power at the switch input in a linear manner. A second type of RF switch compression occurs when the source and drain of the shunting transistor M27 break down at excessive voltages. For submicron silicon-on-insulator (SOI) devices, this voltage may be approximately only +1 VDC above the supply voltage. At breakdown, the shunt device begins to heavily conduct current thereby reducing the power available at the output.
FIG. 2 shows a simplified schematic of a prior art single pole double throw (SPDT) RF switch 20. As shown in FIG. 2, the prior art RF switch 20 minimally includes four MOSFET transistors 23, 24, 27 and 28. The transistors 23 and 24 act as xe2x80x9cpassxe2x80x9d or xe2x80x9cswitchingxe2x80x9d transistors (similar to the switching MOSFET transistor M15 of FIGS. 1a-1c), and are configured to alternatively couple their associated and respective RF input nodes to a common RF node 25. For example, when enabled (or switched xe2x80x9conxe2x80x9d), the switching transistor 23 couples a first RF input signal xe2x80x9cRF1xe2x80x9d, input to a first RF input node 21, to the RF common node 25. Similarly, when enabled, the switching transistor 24 couples a second RF input signal xe2x80x9cRF2xe2x80x9d, input to a second RF input node 22, to the RF common node 25. The shunting transistors, 27 and 28, when enabled, act to alternatively shunt their associated and respective RF input nodes to ground when their associated RF input nodes are uncoupled from the RF common node 25 (i.e., when the switching transistor (23 or 24) connected to the associated input node is turned off).
As shown in FIG. 2, two control voltages are used to control the operation of the prior art RF switch. The control voltages, labeled xe2x80x9cSWxe2x80x9d, and its inverse xe2x80x9cSW_xe2x80x9d, control the operation of the transistors 23, 24, 27 and 28. The control voltages are arranged to alternatively enable (turn on) and disable (turn off) selective transistor pairs. For example, as shown in FIG. 2, when SW is on (in some embodiments this is determined by the control voltage SW being set to a logical xe2x80x9chighxe2x80x9d voltage level, e.g., xe2x80x9c+Vddxe2x80x9d), the switching transistor 23 is enabled, and its associated shunting transistor 28 is also enabled. However, because the inverse of SW, SW_, controls the operation of the second switching transistor 24, and its associated shunting transistor 27, and the control signal SW_ is off during the time period that SW is on (in some embodiments this is determined by SW_ being set to a xe2x88x92Vdd value), those two transistors are disabled, or turned off, during this same time period. In this state (SW xe2x80x9conxe2x80x9d and SW_ xe2x80x9coffxe2x80x9d), the RF1 input signal is coupled to the RF common port 25 (through the enabled switching transistor 23). Because the second switching transistor 24 is turned off, the RF2 input signal is blocked from the RF common port 25. Moreover, the RF2 input signal is further isolated from the RF common port because it is shunted to ground through the enabled shunting transistor 28. As those skilled in the transistor designs arts shall easily recognize, the RF2 signal is coupled to the RF common port 25 (and the RF1 signal is blocked and shunted to ground) in a similar manner when the SW control signal is xe2x80x9coffxe2x80x9d (and SW_ is xe2x80x9conxe2x80x9d).
With varying performance results, RF switches, such as the SPDT RF switch 20 of FIG. 2, have heretofore been implemented in different component technologies, including bulk complementary-metal-oxide-semiconductor (CMOS) and gallium-arsenide (GaAs) technologies. In fact, most high performance high-frequency switches use GaAs technology. The prior art RF switch implementations attempt to improve the RF switch performance characteristics described above, however, they do so with mixed results and with varying degrees of integrated circuit complexity and yields. For example, bulk CMOS RF switches disadvantageously exhibit high insertion loss, low compression, and poor linearity performance characteristics. In contrast, due to the semi-insulating nature of GaAs material, parasitic substrate resistances can be greatly reduced thereby reducing RF switch insertion loss. Similarly, the semi-insulating GaAs substrate improves switch isolation.
Although GaAs RF switch implementations offer improved performance characteristics, the technology has several disadvantages. For example, GaAs technology exhibits relatively low yields of properly functioning integrated circuits. GaAs RF switches tend to be relatively expensive to design and manufacture. In addition, although GaAs switches exhibit improved insertion loss characteristics as described above, they may have low frequency limitations due to slow states present in the GaAs substrate. The technology also does not lend itself to high levels of integration, which requires that digital control circuitry associated with the RF switch be implemented xe2x80x9coff chipxe2x80x9d from the switch. The low power control circuitry associated with the switch has proven difficult to integrate. This is disadvantageous as it both increases the overall system cost or manufacture, size and complexity, as well as reducing system throughput speeds.
It is therefore desirable to provide an RF switch and method for switching RF signals having improved performance characteristics. Specifically, it is desirable to provide an RF switch having improved insertion loss, isolation, and compression. It is desirable that such an RF switch be easily designed and manufactured, relatively inexpensive to manufacture, lend itself to high levels of integration, with low-to-high frequency application. Power control circuitry should be easily integrated on-chip together with the switch functions. Such integration has been heretofore difficult to achieve using Si and GaAs substrates. The present invention provides such an RF switch and method for switching RF signals.
A novel RF switch circuit and method for switching RF signals is described. The RF switch circuit may be used in wireless applications, and may be fabricated in a silicon-on-insulator technology. In one embodiment the RF switch is fabricated on an Ultra-Thin-Silicon (xe2x80x9cUTSixe2x80x9d) substrate. In one embodiment the RF switch includes: an input for receiving an RF signal; a first switching transistor grouping connected to the input to receive the RF signal and connected to an RF common port, wherein the first switching transistor is controlled by a switching voltage (SW); a second switching transistor grouping connected to the first switching transistor grouping and the RF common port, wherein the second switching transistor is controlled by a switching voltage SW_, and wherein SW_ is the inverse of SW so that when the first switching transistor grouping is on, the second switching transistor grouping is off. The switching transistor groupings, when enabled, alternatively connect their respective RF input signals to the RF common port. In this embodiment the RF switch also includes shunting transistor groupings coupled to the switching transistor groupings and also controlled by the switching voltages SW and SW_. The shunting transistor groupings, when enabled, act to alternatively shunt their associated RF input nodes to ground thereby improving RF switch isolation.
The switching and shunting transistor groupings comprise one or more MOSFET transistors connected together in a xe2x80x9cstackedxe2x80x9d or serial configuration. Within each transistor grouping, the gates of the stacked transistors are commonly controlled by a switching voltage (SW or SW_) that is coupled to each transistor gate through respective gate resistors. The stacking of transistor grouping devices and gate resistors increases the compression point of the switch. The RC time constant formed by the gate resistors and the gate capacitance of the MOSFETs is designed to be much longer than the period of the RF signal, causing the RF voltage to be shared equally across the series connected devices. This configuration increases the 1 dB compression point of the RF switch.
A fully integrated RF switch is described that includes digital switch control logic and a negative power supply voltage generator circuit integrated together with the inventive RF switch. In one embodiment, the fully integrated RF switch provides several functions not present in prior art RF switches. For example, in one embodiment, the fully integrated RF switch includes a built-in oscillator that provides clocking input signals to a charge pump circuit, an integrated charge pump circuit that generates the negative power supply voltages required by the other RF switch circuits, CMOS logic circuitry that generates control signals to control the RF switch transistors, level-shifting and low current voltage divider circuits that provide increased reliability of the switch devices, and an RF buffer circuit that isolates RF signal energy from the charge pump and digital control logic circuits. Several embodiments of the charge pump, level shifting, voltage divider, and RF buffer circuits are described. The inventive RF switch provides improvements in insertion loss, switch isolation, and switch compression. In addition, owing to the higher levels of integration made available by the present inventive RF switch, RF system design and fabrication costs are reduced and reliability is increased using the present invention.